Medical image data often face the problem of data scarcity and costly annotation processes. To overcome this, our study introduces a novel transfer learning method for medical image classification. We present a multimodal learning framework that incorporates the pre-trained PubMedCLIP model and multimodal feature fusion. Prompts of different complexities are combined with images as inputs to the proposed model. Our findings demonstrate that this approach significantly enhances image classification tasks while reducing the burden of annotation costs. Our study underscores the potential of PubMedCLIP in revolutionizing medical image analysis through its prompt-based approach and showcases the value of multi-modality for training robust models in healthcare. Code is available at: https://github.com/HongJapan/MTL_prompt_medical.git .
Transfer learning is a key solution to deal with the problem of data scarcity, where learning process leverages knowledge learned from similar tasks. In this study, we employ PubMedCLIP as a pre-trained model for the medical image classification tasks. We evaluate multiple datasets of different body regions rather simultaneously. Moreover, we investigate different amounts of training data to evaluate the effectiveness of PubMedCLIP in the case of limited training data. The experimental results show that transfer learning using PubMedCLIP can achieve good results in across a diverse range of body-region datasets. To the best of our knowledge, this is the first study that employs PubMedCLIP for transfer learning.
Stable periodic-frequent itemset mining is essential in big data analytics with many real-world applications. It involves extracting all itemsets exhibiting stable periodic behaviors in a temporal database. Most previous studies focused on finding these itemsets in row (temporal) databases and disregarded the occurrences of these itemsets in columnar databases. Furthermore, the naïve approach of transforming a columnar database into a row database and then applying the existing algorithms to find interesting itemsets is not practicable due to computational reasons. With this motivation, this paper proposes a framework to discover stable periodic-frequent itemsets in columnar databases. Our framework employs a novel depth-first search algorithm that compresses a given columnar database into a unified dictionary and mines it recursively to find all stable periodic-frequent itemsets. The dictionary holds the information pertaining to itemsets and their temporal occurrences in a database. Experimental results on six databases demonstrate that the proposed algorithm is computationally efficient and scalable.
The rapid spread of rumors on social media, especially during crises like the COVID-19 pandemic, highlights an urgent need for advanced tools to detect fake news. Large Language Models (LLMs), with their vast knowledge and emergent abilities, show great promise in tackling this challenge. This study investigates five state-of-the-art LLMs — DeepSeek, GPT-3.5, GPT-4, Gemini, and Claude— in creating and detecting fake news. Especially, we consider not only the performance but also the similarity of the models. Regarding tweet generation, we identify interesting patterns about similarity among the outputs of the models. Specifically, the outputs of DeepSeek are most similar to those of GPT models, while Gemini’s outputs are more distinct from the others. Regarding detection, no model demonstrates strong performance, even on its own generated dataset. Still, the decisions of DeepSeek are most similar to those of GPT-3.5, while the decisions of Gemini are the least similar to those of the other models. This study provides valuable insights into the dual role of LLMs as both detectors and potential sources of rumors, contributing to the development of more robust and reliable fake news detection systems.